A daughter that spent some of her growing-up years in Baton Rouge sent me the article about cancer in Reserve, and I have relevant, scientific experience on the subject.

Beginning in the 1980s, I did cancer research that involved Louisiana, and that state proved most remarkable. I was working on what are called mathematical partitioning problems; i.e., seeking to divide in the best way possible a set of objects into subsets, so things in one subset are all rather similar – a cluster – and any two clusters are rather dissimilar. There must be a measure of distance between any two objects, and that is what provides the notion of similarity.

I was a faculty member in New York State, and I decided it would be interesting and potentially useful to apply my methods to cancer mortality data and partition our states into clusters, so that if we determined that states X, Y, Z, U, and V constituted a cluster, that would mean they were all similar as to their patterns of death by cancer and that two different clusters would be generally dissimilar.

Such results could be useful towards making hypotheses as to what causes various kinds of cancer. For example, if one wants to study potential environmental causes, he might begin by asking, “What are similar environmental characteristics in all the states, X, Y, …V?” If he wishes to postulate possible genetic causes, he might ask if these states all have numbers of people of similar nationality.

Naturally, if one is interested in what causes cancers he should use incidence data, because not everyone that contracts a disease perishes from it, but incidence data is difficult to collect, whereas cause of death is routinely reported. Therefore, mortality data is a common surrogate.

I was able to find U.S. cancer mortality data for the years 1950 – 1967, already rather dated when I commenced, and it included 50 different tumor types. For my work, a state was represented by a set of 50 numbers, each a measure of the death in that state from one of those cancer types. There being no obvious, correct way to reckon distance between two such objects, I used various, standard methods and made many different computer runs of my clustering method on that data.

What intrigued me most about the results was the persistence of Louisiana NOT to cluster with its geographic, neighbor states, but with Illinois, Minnesota, Michigan, Ohio and the northeast block beginning with Delaware and New Jersey and extending to Canada. My paper appeared in Computers and Biomedical Research, vol. 19, pp. 117-141, 1986, by which time I was a member of the LSU faculty and lived in Baton Rouge.

That was the most popular research I ever accomplished. I never had so many requests for copies of a paper, a local radio station interviewed me, and I received mail from LA residents that wondered if they lived in places that were dangerous as regarded cancer.

By then I had begun my experiments on Louisiana parishes. The state has significant numbers of whites, blacks, females, and males, so I made separate computations on each of the four sub-populations of deceased, because sex and race differences in cancer occurrence are well known.

None of that work ever found its way into print, but over several years I gave talks on it at conferences and in invited lectures at San Francisco; Melbourne, Australia; Bergen, Norway; and Warsaw, Poland.

With the parishes, there were many results that stimulated curiosity and raised questions. For example, in one experiment with 1968-78 death data for black males, Catahoula parish was unlike any other; it was a singleton, a cluster by itself, and showed such radical results as double the rate of stomach cancer deaths; triple, bladder; 7 times for multiple myeloma; 2.5 times for lymphoid leukemia; 3 times, other leukemia; 20 times for floor of mouth (a rare variety); and 8 times for Hodgkin’s Disease.

For white females, 1968-78 deaths, one outcome found the geographically- proximous, northwest parishes, Claiborne, Bienville, DeSoto, Winn, and Grant, joined by Jefferson in the southeast, in a cluster with higher death rates in 17 of the top 20 most mortal, 12 of the second 20, and between 1.5 and 2.2 times the average rate for breast, lung, ovary, cervix, and bladder cancer deaths. In this same experiment Red River, Richland, and E. Carroll parishes, all close in the north of the state, formed a cluster with mortality rates from 1.25 up to 6 times the average for colon, pancreas, ovary, cervix, stomach, secondary respiratory, lymphatic leukemia, brain, larynx, bone, and carcinoma of digestive organs.

For white male cancer deaths in 1968-78, one experiment found Pointe Coupe’e Parish dissimilar to all others with higher than average mortality rates in 17 of the 50 categories but generally much lower than average except for colon, pancreas, and bladder. In this same computer run, a rather linear string of parishes, W. Feliciana, E. Baton Rouge, St. John the Baptist, St. Charles, and Plaquemines was joined by E. Carroll, Cameron, and Vernon to form a cluster that had death rates lower in 19 of the 20 most fatal disease types and higher in but 9 of all the cancer varieties.

To reiterate, a reasonable next step in such work is to examine a cluster, attempt to find similarities among the parishes in it, and examine those with the objective of making reasonable judgments about the causes of the results; i.e., what might account for those patterns of cancer deaths.

Sometime in 1984-1988, several of us faculty at LSU paid a visit to the State Department of Public Health to explore the opportunities of its funding research of this nature. We discovered quickly that there were none; the Department told us it had concluded that Louisiana’s cancer oddities were consequences of the peculiar diet the state’s people ate, and that was that. In 1988, I moved to the University of Alabama and pursued other research.

Notice that this was a long time ago and that the data I had have become old. Recent publications suggest the old Louisiana cancer problems persist and that new research efforts like mine are in order.

Dr. Larry Stanfel
Roundup, Montana
stanfel@midrivers.com